This makes it push the reports into a consistent location if CI_REPORTS is not set.
… runs. What's that all about? At a guess, rm_f doesn't actually take globs.
* feature/cucumber-integration: The remainder of the specs, I think. Not test-driving is painful... Big pile o' passing specs. I suppose it's testing that it interacts correctly with the underlying API... Spec out the CucumberFailure. Nuke the CucumberDoc version for now. Inherit directly from Ast::Visitor. Accept (and demonstrate) that Cucumber doesn't differentiate between failures and errors. Implement catching and reporting of errors. If there are 0 errors, don't check for errors elements. Reimplement the visitor. I must have been smoking something on Friday. Pending steps should not be reported as errors. Start of an implementation of the visitor. Expect a slightly different filename. We have to require the cucumber loader slightly differently otherwise it doesn't autoload step definitions. Acceptance tests for cucumber integration. Add in a place to stick specs for the Cucumber integration. Add in some tests for the Cucumber setup rake task. Infrastructure for Cucumber integration.
…orrectly with the underlying API...
…failures and errors.
Now at least it actually creates testcases too, instead of erroneously creating new test suites. What a numpty.
… it doesn't autoload step definitions.
Wow. People test their Rakefiles? Never thought to do that before...
Don't just have a generic rescue block. By default, a paramater-less rescue clause will only catch exceptions that inherit from StandardError. Gem::LoadError doesn't, which means it would never be caught by a blank rescue clause. Signed-off-by: Nick Sieger <email@example.com>